Unwanted Traffic: Denial of Service Attacks · PDF file5 Smurf amplification DoS attack Send ping request to broadcast addr (ICMP Echo Req) Lots of responses: n Every host on target

1

Unwanted Traffic:Denial of Service Attacks

Dan Boneh

CS 155

2

What is network DoS?Goal: take out a large site with little computing work

How: Amplificationn Small number of packets ⇒ big effect

Two types of amplification attacks:n DoS bug:

w Design flaw allowing one machine to disrupt a service

n DoS flood:w Command bot-net to generate flood of requests

3

DoS can happen at any layerThis lecture:

n Sample Dos at different layers (by order):w Linkw TCP/UDPw Application

n DoS mitigations

Sad truth: n Current Internet not designed to handle DDoS attacks

4

Warm up: 802.11b DoS bugsRadio jamming attacks: trivial, not our focus.

Protocol DoS bugs: [Bellardo, Savage, ’03]

n NAV (Network Allocation Vector):w 15-bit field. Max value: 32767w Any node can reserve channel for NAV secondsw No one else should transmit during NAV period

… but not followed by most 802.11b cards

n De-authentication bug:w Any node can send deauth packet to APw Deauth packet unauthenticated

⇒ attacker can repeatedly deauth anyone

5

Smurf amplification DoS attack

Send ping request to broadcast addr (ICMP Echo Req) Lots of responses:n Every host on target network generates a ping

reply (ICMP Echo Reply) to victim

Prevention: reject external packets to broadcast address

gatewayDoSSource

DoSTarget

1 ICMP Echo ReqSrc: Dos TargetDest: brdct addr

3 ICMP Echo ReplyDest: Dos Target

6

Modern day example (Mar ’13)

2006: 0.58M open resolvers on Internet (Kaminsky-Shiffman)2017: 15M open resolvers (openresolverproject.org)

⇒ 3/2013: DDoS attack generating 309 Gbps for 28 mins.

DNSServer

DoSSource

DoSTarget

DNS QuerySrcIP: Dos Target

(60 bytes)EDNS Reponse

(3000 bytes)

DNS Amplification attack: ( ×50 amplification )

13

By way of contrast, 76 percent of respondents (Figure 12) indicated that the purported geopolitical origin of traffic ingressingand traversing their networks has a significant impact on their perception of the threat that this traffic may pose to theirorganization and/or end customers.

Scale, Targeting and Frequency of Attacks

As illustrated in Figure 1 (page 5) and again in Figure 13, the highest-bandwidth attack observed by respondents during thesurvey period was a 100 Gbps DNS reflection/amplification attack. This represents a 102 percent increase over the previousyear. It is also the single largest increase in attack bandwidth year over year since the first report in 2005 and a 1000 percentincrease in attack bandwidth since the report’s inception.

Based upon our experiences working with operators over the last year, we believe this large increase in attack-traffic bandwidth maybe partially due to operators focusing their defenses against lower-bandwidth and application-layer DDoS attacks. Attackers mayhave had to “up the ante” to overwhelm the defenses and bandwidth capacity of defenders. Additionally, the increased availability ofbotted hosts, combined with the growing popularity of DNS amplification/reflection attacks, has also played a role in this escalation.

Worldwide Infrastructure Security Report, Volume VI

Ban

dwid

th(G

bps)

2005 2006 2007 2008 2009 2010

Scale, Targeting and Frequency of Attacks100

90

80

70

60

50

40

30

20

10

0

100 Gbps

Figure 13Source: Arbor Networks, Inc.

Source: Arbor Networks, Inc.

Influence of Geopolitical Origin of Network Traffic on Threat Perception

Influential

Not Influential

76%

24%

Figure 12Source: Arbor Networks, Inc.

7Feb.2014:400Gbps viaNTPamplification(4500NTPservers)

8

Review: IP Header formatConnectionlessn Unreliablen Best effort

Version Header LengthType of Service

Total LengthIdentification

Flags

Time to LiveProtocol

Header Checksum

Source Address of Originating Host

Destination Address of Target Host

Options

Padding

IP Data

Fragment Offset

0 31

9

Review: TCP Header formatTCP:n Session basedn Congestion controln In order delivery

Source Port Dest portSEQ NumberACK Number

Other stuff

URG

PSR

ACK

PSH

SYN

FIN

0 31

10

Review: TCP Handshake

C S

SYN:

SYN/ACK:

ACK:

Listening

Store SNC , SNS

Wait

Established

SNC⟵randCANC⟵0

SNS⟵randSANS⟵SNC

SN⟵SNCAN⟵SNS

11

TCP SYN Flood I: low rate (DoS bug)

C

SYNC1

SYNC2

SYNC3

SYNC4

SYNC5

S Single machine:• SYN Packets with

random source IPaddresses

• Fills up backlog queueon server

• No further connectionspossible

12

SYN Floods (phrack 48, no 13, 1996)

OSBacklog

queue sizeLinux 1.2.x 10FreeBSD 2.1.5 128WinNT 4.0 6

Backlog timeout: 3 minutes

• Attackerneedsonly128SYNpacketsevery3minutes• LowrateSYNflood

13

A classic SYN flood example

MS Blaster worm (2003)n Infected machines at noon on Aug 16th:

w SYN flood on port 80 to windowsupdate.comw 50 SYN packets every second.

n each packet is 40 bytes.w Spoofed source IP: a.b.X.Y where X,Y random.

MS solution: n new name: windowsupdate.microsoft.com

14

Low rate SYN flood defenses

Non-solution:n Increase backlog queue size or decrease timeout

Correct solution (when under attack) : n Syncookies: remove state from servern Small performance overhead

15

SyncookiesIdea: use secret key and data in packet to gen. server SN

Server responds to Client with SYN-ACK cookie:n T = 5-bit counter incremented every 64 secs.

n L = MACkey (SAddr, SPort, DAddr, DPort, SNC, T) [24 bits]

w key: picked at random during boot

n SNS = (T . mss . L) ( |L| = 24 bits )

n Server does not save state (other TCP options are lost)

Honest client responds with ACK ( AN=SNS , SN=SNC+1 )n Server allocates space for socket only if valid SNS

[Bernstein, Schenk]

16

SYN floods: backscatter [MVS’01]

SYN with forged source IP Þ SYN/ACK to random host

17

Backscatter measurementListen to unused IP addresss space (darknet)

Lonely SYN/ACK packet likely to be result of SYN attack

2001: 400 SYN attacks/week2013: 773 SYN attacks/24 hours (arbor networks ATLAS)

n Larger experiments: (monitor many ISP darknets)w Arbor networks

0 232monitor/8 network

Estonia attack (ATLAS ‘07)

Attack types detected: n 115 ICMP floods, 4 TCP SYN floods

Bandwidth:n 12 attacks: 70-95 Mbps for over 10 hours

All attack traffic was coming from outside Estonian Estonia’s solution:

w Estonian ISPs blocked all foreign traffic until attacks stopped⇒ DoS attack had little impact inside Estonia

18

19

Massive floods

Command bot army to flood specific target: (DDoS)

• Flood with SYN, ACK, UDP, and GRE packets

• 623 Gbps (peak) from ≈100K compromised IoT devices

• At web site:• Saturates network uplink or network router• Random source IP ⇒

attack SYNs look the same as real SYNs

• What to do ???

(e.g. Mirai 9/2016 on Krebs)

20src: incapsula.com

21

Google project shieldProtecting news organizations.

(Commercial service: Akamai, Cloudlare, … )

Idea: only forward established TCP connections to site

ProjectShieldProxy

Web site

Lots-of-SYNs

Lots-of-SYN/ACKs

Few ACKsForwardto site

22

Other junk packets

Proxy must keep floods of these away from web site

Attack Packet Victim Response Rate: attk/day[ATLAS 2013]

TCP SYN to open port TCP SYN/ACK 773

TCP SYN to closed port TCP RST

TCP ACK or TCP DATA TCP RST

TCP RST No response

TCP NULL TCP RST

ICMP ECHO Request ICMP ECHO Response 50

UDP to closed port ICMP Port unreachable 387

23

Stronger attacks: GET floodCommand bot army to:

n Complete TCP connection to web siten Send short HTTP GET requestn Repeat

Will bypass SYN flood protection proxy

… but:n Attacker can no longer use random source IPs.

w Reveals location of bot zombies

n Proxy can now block or rate-limit bots.

A real-world example: GitHub (3/2015)

Javascript-based DDoS:

24

function imgflood() { var TARGET = 'victim-website.com/index.php?’var rand = Math.floor(Math.random() * 1000)var pic = new Image()pic.src = 'http://'+TARGET+rand+'=val'

}setInterval(imgflood, 10)

imageFlood.js

github.comhonest end user

popularserver

injectimageFlood.js

Would HTTPS prevent this DDoS?

25

DNS DoS Attacks (e.g. Dyn attack 10/2016)

DNS runs on UDP port 53• DNS entry for victim.com hosted at DNSProvider.com

DDoS attack:• flood DNSProvider.com with DNS queries• Random source IP address in UDP packets• Takes out entire DNS server (collateral damage)

Dyn attack: used some Mirai-based bots• At least 100,000 malicious end points

⇒ Dyn cannot answer many legit DNS queries ⇒ Disrupted service at Netflix, Github, Twitter, …

DoS via route hijacking YouTube is 208.65.152.0/22 (includes 210 IP addr)

youtube.com is 208.65.153.238, …

Feb. 2008:n Pakistan telecom advertised a BGP path for

208.65.153.0/24 (includes 28 IP addr)n Routing decisions use most specific prefixn The entire Internet now thinks

208.65.153.238 is in Pakistan

Outage resolved within two hours… but demonstrates huge DoS vuln. with no solution!

26

27

DoS at higher layersSSL/TLS handshake [SD’03]

n RSA-encrypt speed ≈ 10 × RSA-decrypt speed⇒ Single machine can bring down ten web servers

Similar problem with application DoS:n Send HTTP request for some large PDF filen Easy work for client, hard work for server.

WebServer

Client Hello

Server Hello (pub-key)

Client key exchangeRSAEncrypt RSA

Decrypt

28

DoS Mitigation

29

1. Client puzzlesIdea: slow down attacker

Moderately hard problem:n Given challenge C find X such that

LSBn ( SHA-1( C || X ) ) = 0n

n Assumption: takes expected 2n time to solven For n=16 takes about .3sec on 1GhZ machinen Main point: checking puzzle solution is easy.

During DoS attack:n Everyone must submit puzzle solution with requestsn When no attack: do not require puzzle solution

30

Examples

GET floods (RSA ‘99)n Example challenge: C = TCP server-seq-numn First data packet must contain puzzle solution

w Otherwise TCP connection is closed

SSL handshake DoS: (SD’03)n Challenge C based on TLS session IDn Server: check puzzle solution before RSA decrypt.

31

Benefits and limitations

Hardness of challenge: nn Decided based on DoS attack volume.

Limitations:

n Requires changes to both clients and servers

n Hurts low power legitimate clients during attack:w Clients on cell phones and tablets cannot connect

32

Memory-bound functionsCPU power ratio:n high end server / low end cell phone = 8000⇒ Impossible to scale to hard puzzles

Interesting observation:n Main memory access time ratio:

w high end server / low end cell phone = 2

Better puzzles:n Solution requires many main memory accesses

w Dwork-Goldberg-Naor, Crypto ‘03w Abadi-Burrows-Manasse-Wobber, ACM ToIT ‘05

33

2. CAPTCHAsIdea: verify that connection is from a human

Applies to application layer DDoS [Killbots ’05]n During attack: generate CAPTCHAs and process

request only if valid solutionn Present one CAPTCHA per source IP address.

34

3. Source identificationGoal: identify packet source

Ultimate goal: block attack at the source

35

1. Ingress filtering (RFC 2827, 3704)

Big problem: DDoS with spoofed source IPs

Ingress filtering policy: ISP only forwards packets with legitimate source IP (see also SAVE protocol)

ISP Internet

Implementation problemsALL ISPs must do this. Requires global trust.

n If 10% of ISPs do not implement ⇒ no defensen No incentive for deployment

2017: n 33% of Auto. Systems are fully spoofable

(spoofer.caida.org)n 23% of announced IP address space is spoofable

Recall: 309 Gbps attack used only 3 networks (3/2013)

37

2. Traceback [Savage et al. ’00]

Goal:n Given set of attack packetsn Determine path to source

How: change routers to record info in packets

Assumptions:n Most routers remain uncompromisedn Attacker sends many packets n Route from attacker to victim remains relatively

stable

38

Simple methodWrite path into network packetn Each router adds its own IP address to packetn Victim reads path from packet

Problem:n Requires space in packet

w Path can be longw No extra fields in current IP format

n Changes to packet format too much to expect

39

Better ideaDDoS involves many packets on same path

Store one link in each packetn Each router

probabilistically stores own address

n Fixed space regardless of path length

R6 R7 R8

A4 A5A1 A2 A3

R9 R10

R12

V

40

Edge SamplingData fields written to packet:n Edge: start and end IP addressesn Distance: number of hops since edge stored

Marking procedure for router Rif coin turns up heads (with probability p) then

write R into start addresswrite 0 into distance field

elseif distance == 0 write R into end fieldincrement distance field

41

Edge Sampling: picturePacket receivedn R1 receives packet from source or another routern Packet contains space for start, end, distance

R1 R2 R3

packet s e d

42

Edge Sampling: pictureBegin writing edgen R1 chooses to write start of edgen Sets distance to 0

R1 R2 R3

packet R1 0

43

Edge Sampling

packet R1 R2 1

R1 R2 R3

Finish writing edgen R2 chooses not to overwrite edgen Distance is 0

w Write end of edge, increment distance to 1

44

Edge Sampling

packet R1 R2 2

R1 R2 R3

Increment distancen R3 chooses not to overwrite edgen Distance >0

w Increment distance to 2

45

Path reconstructionExtract information from attack packets

Build graph rooted at victimn Each (start,end,distance) tuple provides an edge

# packets needed to reconstruct path

E(X) <

where p is marking probability, d is length of path

ln(d) p(1-p)d-1

46

Details: where to store edgeIdentification fieldn Used for fragmentationn Fragmentation is raren 16 bits

Store edge in 16 bits?

n Break into chunksn Store start + end

Version Header LengthType of Service

Total LengthIdentification

Flags

Time to LiveProtocol

Header Checksum

Source Address of Originating Host

Destination Address of Target Host

Options

Padding

IP Data

Fragment OffsetIdentification

offset distance edge chunk0 2 3 7 8 15

47

More traceback proposalsAdvanced and Authenticated Marking Schemes for IP Tracebackn Song, Perrig. IEEE Infocomm ’01n Reduces noisy data and time to reconstruct paths

An algebraic approach to IP tracebackn Stubblefield, Dean, Franklin. NDSS ’02

Hash-Based IP Traceback n Snoeren, Partridge, Sanchez, Jones, Tchakountio,

Kent, Strayer. SIGCOMM ‘01

48

Problem: Reflector attacks [Paxson ’01]

Reflector: n A network component that responds to packetsn Response sent to victim (spoofed source IP)

Examples:n DNS Resolvers: UDP 53 with victim.com source

w At victim: DNS response

n Web servers: TCP SYN 80 with victim.com sourcew At victim: TCP SYN ACK packet

n NTP servers

49

DoS AttackSingle Master

Many bots to generate flood

Zillions of reflectors to hide botsn Kills traceback and

pushback methods

50

Capability based defense

51

Capability based defenseAnderson, Roscoe, Wetherall. n Preventing internet denial-of-service with

capabilities. SIGCOMM ‘04.

Yaar, Perrig, and Song. n Siff: A stateless internet flow filter to mitigate DDoS

flooding attacks. IEEE S&P ’04.

Yang, Wetherall, Anderson. n A DoS-limiting network architecture.

SIGCOMM ’05

52

Capability based defenseBasic idea:n Receivers can specify what packets they want

How:n Sender requests capability in SYN packet

w Path identifier used to limit # reqs from one sourcen Receiver responds with capabilityn Sender includes capability in all future packets

n Main point: Routers only forward:w Request packets, andw Packets with valid capability

53

Capability based defenseCapabilities can be revoked if source is attackingn Blocks attack packets close to source

R1R2

R3 R4dest

Source AS Transit AS Dest AS

Attack packets dropped

54

Pushback Traffic Filtering

55

Pushback filteringMahajan, Bellovin, Floyd, Ioannidis, Paxson, Shenker. Controlling High Bandwidth Aggregates in the Network. Computer Communications Review ‘02.

Ioannidis, Bellovin. Implementing Pushback: Router-Based Defense Against DoS Attacks. NDSS ’02

Argyraki, Cheriton. Active Internet Traffic Filtering: Real-Time Response to Denial-of-Service Attacks. USENIX ‘05.

56

Pushback Traffic FilteringAssumption: DoS attack from few sources

Iteratively block attacking network segments.

57

Overlay filtering

58

Overlay filteringKeromytis, Misra, Rubenstein. SOS: Secure Overlay Services. SIGCOMM ‘02.

D. Andersen. Mayday.Distributed Filtering for Internet Services.Usenix USITS ‘03.

Lakshminarayanan, Adkins, Perrig, Stoica.Taming IP Packet Flooding Attacks. HotNets ’03.

59

Take home message:Denial of Service attacks are real:

Must be considered at design time

Sad truth: n Internet is ill-equipped to handle DDoS attacks n Many commercial solutions: CloudFlare, Akamai, …

Many proposals for core redesign

60

THE END